iHore/vLLM
8
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[XPU] Ensure RNG offset alignment with PyTorch requirements in XPU sampler (#43028)

Browse Source 
Signed-off-by: chaojun-zhang <chaojun.zhang@intel.com>
Signed-off-by: Chaojun Zhang <chaojun.zhang@intel.com>
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
Co-authored-by: mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
This commit is contained in:
Chaojun Zhang
2026-05-26 02:01:30 +00:00
committed by GitHub
co-authored by gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com> mergify[bot] <37929162+mergify[bot]@users.noreply.github.com>
parent d4004455d2
commit 71d810bbf4
1 changed files with 2 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files
vllm/v1/sample/ops/topk_topp_sampler.py
+2 -1
View File
@@ -291,7 +291,8 @@ class TopKTopPSampler(nn.Module):
)
# The custom XPU sampler kernel consumes RNG values internally, so advance
# the default generator's offset to keep future draws deterministic.
offset += logits.numel()
# pytorch: offset must be multiple of 4
offset = (offset + logits.numel() + 3) // 4 * 4
state.view(torch.int64)[1] = offset
generator.set_state(state)
return random_sampled, logits_to_return